Abstract: To improve the accuracy of climate models in precipitation simulation, we collect the monthly precipitation at 12 meteorological stations over the Yellow River basin upstream of Tangnaihai, and correct the deviation in precipitation simulated by 18 global climate models (GCMs) using three methods, i.e., single gamma (SG), separated double gamma (SDG), and separated triple gamma (STG) distributions. The effects of these bias-correction methods are evaluated and compared in terms of three indexes: bias-correction effective index C, Nash-Sutcliffe (NS) efficiency coefficient, and relative bias index (PBIAS). The results indicate that all the three methods are able to effectively remove the deviation in simulated precipitations from the observed ones. After correction, 69.4%, 82% and 88.4% of index C values exceed 0.85 for SG, SDG and STG, respectively; 69%, 70.8% and 71.3% of NS values exceed 0.4; 65.7%, 67.6% and 69.9% of index PBIAS values are -1.3% to 2.3%, a range close to zero. In summary, STG is the best in precipitation bias correction and SG is the worst.